Identifying and assessing critical uncertainty thresholds in a forest pest risk model
Pest risk maps can provide helpful decision support for invasive alien species management, but often fail to address adequately the uncertainty associated with their predicted risk values. Th is chapter explores how increased uncertainty in a risk model’s numeric assumptions (i.e. its principal parameters) might aff ect the resulting risk map. We used a spatial stochastic model, integrating components for entry, establishment and spread, to estimate the risks of invasion and their variation across a two-dimensional gridded landscape for Sirex noctilio, a non-native woodwasp detected in eastern North America in 2004. Historically, S. noctilio has been a major pest of pine (Pinus spp.) plantations in the southern hemisphere. We present a sensitivity analysis of the mapped risk estimates to variation in six key model parameters: (i) the annual probabilities of new S. noctilio entries at US and Canadian ports; (ii) the S. noctilio population-carrying capacity at a given location; (iii) the maximum annual spread distance; (iv) the probability of local dispersal (i.e. at a distance of 1 km); (v) the susceptibility of the host resource; and (vi) the growth rate of the host trees. We used Monte Carlo simulation to sample values from symmetric uniform distributions defi ned by a series of nested variability bounds around each parameter’s initial values (i.e. ±5%, …, ±50%). Th e results show that maximum annual spread distance, which governs longdistance dispersal, was the most sensitive of the tested parameters. At ±15% uncertainty in this parameter, mapped risk values shifted notably. No other parameter had a major eff ect, even at wider bounds of variation. Th e methods presented in this chapter are generic and can be used to assess the impact of uncertainties on the stability of pest risk maps or to identify any geographic areas for which management decisions can be made confi dently, regardless of uncertainty.